THE CONSTRUCT AND CONCURRENT VALIDITY OF WORKER/PEER ATTACHMENT. A Dissertation Presented to The Academic Faculty. Erin Page

Transcription

1 THE CONSTRUCT AND CONCURRENT VALIDITY OF WORKER/PEER ATTACHMENT A Dissertation Presented to The Academic Faculty By Erin Page In Partial Fulfillment Of the Requirements for the Degree Doctor of Philosophy in the School of Psychology Georgia Institute of Technology December 2007

2 The Construct and Concurrent Validity of Worker/Peer Attachment Approved by: Dr. Phillip Ackerman School of Psychology Georgia Institute of Technology Dr. Larry James School of Psychology Georgia Institute of Technology Dr. Ruth Kanfer, Advisor School of Psychology Georgia Institute of Technology Dr. Scott Lilienfeld School of Psychology Emory University Dr. Charles Parsons College of Management Georgia Institute of Technology Date Approved: August 21, 2007

3 ACKNOWLEDGMENTS To the members of my committee, thank you for your time and challenging critiques. The varied frameworks and research traditions you provided gave me great perspective on the field. Dr. Ruth Kanfer, thank you for your exceptional advising and support. Your insistence on high-quality work and thoroughness pushed me to maximize my ability and knowledge. I will benefit from the approach you taught me and your insight into industrial and organizational psychology throughout my career. To Rebecca McAlonan, thank you for all of your hard work and help in making my dissertation project a reality. To all of my friends at Georgia Tech, thank you for your support and friendship throughout my graduate school career. I enjoyed the entire journey, from late-night study sessions to barbequing on the weekends. To my family, Mom, Dad, Brie, Grammy, and Grandad, without your encouragement and help throughout my life I would not be where I am today. iii

4 TABLE OF CONTENTS ACKNOWLEDGMENTS...III LIST OF TABLES... VII LIST OF FIGURES...IX SUMMARY... X CHAPTER 1 INTRODUCTION... 1 Background...5 Overview...5 Adult Attachment...9 Attachment and Work...15 Definition...24 Measurement...26 Previous Techniques...27 Dimensionality...33 Target...36 Reliability Issues...39 Psychometric Properties...40 Test-Retest Reliability...40 Inter-Rater Reliability...41 Antecedents...43 Tenure...48 Construct Validity...49 Trait Complexes...51 Murray's Needs...52 Personality...57 Interpersonal Trust...65 Criterion-Related Validity...68 Coping...69 Work-Family Conflict...74 Performance...79 Adjustment...82 Organizational Commitment...84 Summary Hypotheses...88 iv

5 CHAPTER 2 (STUDY 1) Method...94 Participants...94 Procedure...95 Measures...96 Peer Attachment Scale (PAS)...96 Peer Attachment Scale Peer Rating (PAS PR)...97 Romantic/Close Relationship Attachment...97 Edwards Personal Preference Schedule (EPPS)...97 Big Five IPIP...98 Interpersonal Trust at School...98 Brief COPE...98 School-Family/Life Conflict...99 Student Adaptation to College Questionnaire (SACQ)...99 Organizational Commitment (Short Form) Results Discussion CHAPTER 3 (STUDY 2) Method Participants Procedure Measures WAS Romantic/Close Relationship Attachment Socialization Self-Report Unit Tenure EPPS Big Five IPIP Interpersonal Trust at Work Brief COPE Work-Family Conflict Job Performance Ratings Organizational Commitment Department Results Discussion CHAPTER 4 (GENERAL DISCUSSION) Scale Properties of the WAS and PAS Convergent and Discriminant Validities Attachment Antecedents in the Work Setting Attachment Process in the Work Setting v

6 Sample and Setting Issues Sample Characteristics Work/Tasks Criteria Incremental Predictive Validity Attachment and Coping Attachment Avoidance and Coping Attachment Anxiety and Coping Attachment Anxiety and Family-Work Conflict Attachment, College Adjustment, and Work Performance Attachment Anxiety and College Adjustment Attachment Avoidance and College Adjustment Attachment Anxiety and Performance Attachment Avoidance and Commitment Practical Implications Limitations Future Research Conclusion APPENDIX A APPENDIX B APPENDIX C APPENDIX D APPENDIX E APPENDIX F APPENDIX G REFERENCES vi

7 LIST OF TABLES Table 1. Principal Axis Factor Analysis of the PAS (Hypothesis 1) Table 2. Descriptive Statistics and Reliability Analyses (Hypotheses 1-3) Table 3. Antecedent Analyses of Peer Attachment (Hypothesis 4) Table 4. Descriptive Statistics and Correlations among Correlates (Hypothesis 6) Table 5. Descriptive Statistics and Correlations among Correlates using Average Peer Ratings Table 6. Comparison of Attachment Anxiety and Avoidance Correlates using Peer and Self-Ratings Table 7. Descriptive Statistics and Correlations among Criteria Table 8. Analyses of Peer Attachment Anxiety s Criterion-Related Validity (Hypotheses 7a, 7b, 8a, 10a, and 11a) Table 9. Analyses of Peer Attachment Avoidance s Criterion-Related Validity (Hypotheses 7c, 7d, 8b, 10b, and 11b) Table 10. Analyses of Peer Attachment Anxiety s Incremental Predictive Validity over Actual Correlates Table 11. Analyses of Peer Attachment Avoidance Incremental Predictive Validity over Actual Correlates Table 12. Principal Axis Factor Analysis of the WAS (Hypothesis 1) Table 13. Descriptive Statistics and Antecedent Analyses of Worker Attachment (Hypothesis 4) Table 14. Average Department Tenure s Prediction of Within-Department Attachment Agreement (Hypothesis 5) Table 15. Descriptive Statistics and Correlations among Correlates (Hypothesis 6) Table 16. Comparison of Attachment Anxiety and Avoidance Correlates Table 17. Descriptive Statistics and Correlations among Criteria vii

8 Table 18. Worker Attachment and Total Performance Ratings Table 19. Analyses of Worker Attachment Anxiety s Criterion-Related Validity (Hypotheses 7a, 7b, 8a, 9a, 9b, and 11a) Table 20. Analyses of Worker Attachment Avoidance s Criterion-Related Validity (Hypotheses 7c, 7d, 8b, 9c, 9d, and 11b) Table 21. Worker Attachment Anxiety s Predictive Validity for Work Behaviors Table 22. Analyses of Worker Attachment Avoidance s Incremental Predictive Validity over Actual Correlates Table 23. Criterion-Related Validity Hypotheses Supported in the Current Investigation Table D1. Alternate versus Hypothesized Model Table E1. Participant Relationship with Peer and Peer Ratings Table F1. Descriptive Statistics for Department Tenure and Attachment Homogeneity Table F2. Descriptive Statistics for Workgroup Tenure and Attachment Homogeneity Table F3. Average Workgroup Tenure s Prediction of Within-Workgroup Attachment Agreement Table G1. Worker Attachment and Self Performance Ratings Table G2. Worker Attachment and Manager Performance Ratings Table G3. Worker Attachment and Coworker Performance Ratings Table G4. Interpersonal Performance (360 Degree Ratings) Table G5. Technical Performance (360 Degree Ratings) viii

9 LIST OF FIGURES Figure 1. Dimensional model of adult attachment. Adapted from Brennan, Clark, & Shaver (1998)...35 ix

10 SUMMARY Despite decades of interest in the organizational and managerial factors that influence an individual s attachment to the organization, very little is known about the personal characteristics that may influence an individual s interpersonal relations at work and attachment to the organization. In the context of a changing workplace, in which greater importance is being placed on worker interpersonal skills and organizational commitment, individual differences in adult attachment styles represent a potentially important determinant of intrapersonal and interpersonal criteria as well as organizational commitment. The purpose of this research was to empirically validate a work-related measure of adult attachment and to examine the criterion-related validity of this measure for work outcomes and organizational commitment. The research builds upon social psychological theories and research investigating individual differences in attachment styles during childhood (e.g., Bowlby 1969/1982) and in romantic relationships (e.g., Hazan and Shaver, 1987). Using findings obtained by Brennan, Clark, and Shaver (1998), two versions of a two-dimensional measure of workrelated attachment were developed and investigated in two work-related contexts: academic work and paid employment. The two studies investigated the construct validity of the measures in each setting. In Study 1, conducted in the academic context, evidence on test reliability, discriminant and convergent validity, and the criterion-related validity of the measure for college adjustment was examined. In Study 2, conducted in the work context, evidence was obtained on the construct validity of the measure and the concurrent validity of the measure for a variety of work outcomes related to the process x

11 of organizational attachment, employee well-being and job performance. The overall patterns of findings provide empirical evidence for the psychometric properties, construct validity and criterion-related validity of the new measures across work settings; and indicate the role that individual differences in attachment plays in predicting organizationally-relevant work outcomes, such as coping, work-life balance, adjustment, and organizational commitment. This paper contains background on infant and adult attachment, a definition and explanation of worker attachment, previous and current measurement techniques for attachment, and a framework that demonstrates the construct validity of worker attachment. Many of the correlate variables had existing relationships with adult attachment and therefore their relationships with worker and peer attachment served as construct validation. Yet, other variables, especially the antecedents and criteria, demonstrated new relationships in the current contexts that had not previously been explored (e.g., organizational socialization, organizational commitment and work-life balance). The results of the construct and criterion-related validation studies on worker/peer attachment will be discussed, and in the final section of this paper the result s implications on today s workplace will be considered. xi

12 CHAPTER 1 INTRODUCTION Changes in the nature of work, such as the post-industrial boom of service-based jobs, have spurred greater attention to factors that affect interpersonal relations within the workplace (Mount, Barrick, & Stewart, 1998). The increased use of team structures and leadership strategies designed to promote employee commitment to the work unit and coworkers has led to extensive research on the work conditions and situations that promote employee cooperation in teams and commitment (e.g., Mathieu & Zajac, 1990; Mitchell, Holtom, Lee, & Erez, 2001). To date, however, a lot of research in this area has focused on work conditions. The research on the characteristics of workers that contribute to positive interpersonal relations in the workplace has been disappointing. Existing dispositional predictors of interpersonal performance and commitment, such as affiliation or agreeableness, have not allowed for strong assessment tools (e.g., Meyer, Allen, & Smith, 1993; Mount et al., 1998). These variables tend to be one-dimensional assessments of preferences for social settings and do not get at the dyadic nature of relationships. They measure what types of situations people prefer and don't assess what one expects from and how they interact with other people. In summary, employee relations and organizational commitment to date have been viewed primarily as a result of situational influences such as job insecurity and job redesign. The current investigation builds on existing theories and research on childhood attachment and adult romantic attachment for the purpose of developing measures that capture individual differences in relational tendencies in achievement settings, such as the 1

13 workplace and academia. Attachment theory, in which attachment is defined as the "systematic patterns of expectations, needs, emotions, emotion-regulation, and social behavior" (Shaver & Mikulincer, 2002, p. 134), has been used throughout developmental and social psychology as a framework for studying interpersonal relationships (e.g., Ainsworth, Blehar, Waters, & Wall, 1978; Hazan & Shaver, 1987). In general, Bowlby s (1969) attachment theory describes internal working models that reflect individual differences in terms of self, others and the interpersonal environment. These internal working models are created through early experiences with separation and attachment and are thought to continue into adulthood and reflect distinct behavior patterns (Bowlby, 1969/1982). Attachment has been used to look at infant-primary caregiver relationships, romantic relationship, close relationships, and relationships within groups or teams (Ainsworth et al., 1978; Brennan, Clark, & Shaver, 1998; Hazan & Shaver, 1987; Smith et al., 1999). The research on adult attachment (i.e., romantic and close relationship attachment) has flourished within the social and developmental domains since its introduction (Hazan and Shaver, 1987; Rholes & Simpson, 2004). This research is grounded in childhood/infant attachment theory (Bowlby, 1968/1982), and has focused on construct measurement, specifying the relations among adult attachment and other personality trait measures and adult behavior in close relationships. Adult attachment has been studied by social and personality psychologists using measures of attachment to one's current peers or a romantic partner (i.e., close relationships are viewed as attachment processes), whereas developmental psychologists have focused on attachment within the nuclear family, typically by measuring attachment to a primary caregiver (Bernier & Dozier, 2002). Adult attachment has been associated 2

14 with other non-ability traits (Gallo, Smith, & Ruiz, 2003), but it has also shown unique predictive validity for important social behaviors and attitudes making it an essential prediction tool in development and social psychology (Griffin & Bartholomew, 1994). Existing research on romantic partners has shown that individual differences in attachment significantly predict relationship status, interpersonal dependency, relationship length, commitment, and relationship satisfaction even after other variables such as the Big Five (i.e., a commonly studied theory of personality including 5 personality traits: neuroticism, extraversion, conscientiousness, agreeableness and openness to experience) are statistically controlled. Specifically, the Big Five had a.17 multiple correlation with relationship status, which changed to.33 when attachment variables were added ( R 2 =.08, p <.02). Also, there was a.45 multiple correlation between relationship length and the Big Five, which increased to.61 ( R 2 =.17, p <.001) when attachment was added. Similarly, the prediction of relationship satisfaction increased when attachment variables were added to the Big Five to.49 ( R 2 =.09, F = 2.95, p =.04). Finally, a.13 increase in R 2 to.45 ( F = 4.05, p <.01) was found when attachment variables were added to the Big Five for the prediction of relationship commitment (Griffin & Bartholomew, 1994; Shaver & Brennan, 1992). Using the same strategy of extension used by researchers in the romantic attachment domain, the current research began with Bowlby s (1969/1982) latent construct of attachment as the foundational construct from which to develop individual differences variables that reflects tendencies and preferences with respect to interpersonal relations in the workplace. In the current investigation measures of attachment predicted selective attitudes, coping styles, and behaviors at work pertaining to social relations, and 3

15 organizational commitment (i.e., a work-related analog to relationship commitment). Similar to the theory and methodology used in the adult attachment literature, worker/peer attachment was measured using a self-report questionnaire (e.g., Brennan et al., 1998) and validated by examining its reliability, relations with personality variables, and prediction of work-related attitudes and behaviors. The viability of the measures was investigated in a two-part validation process using two distinct samples: academia and work. This paper is divided into the following sections. In the first section, I review previous research on infant and adult (i.e., romantic or close relationship) attachment. Then, a definition of worker attachment is provided, which is followed by a discussion on previous attachment measurement techniques as well as the current methodology. After outlining the measurement process, the theoretical antecedents to worker/peer attachment are addressed. Then, I discuss the process of attachment within the organization. After the construct has been introduced, there is a section on the construct validity of worker/peer attachment in which theoretically similar correlates (i.e., constructs with previous relationships with close relationship attachment or conceptual relationships with worker/peer attachment) are presented along with their proposed relationships with worker/peer attachment. Then, criteria associated with the new predictors, worker and peer attachment, are reviewed and relationships between the predictors and interpersonal or relationship-based criteria are proposed. At the end of Chapter 1 a summary of all of the hypotheses will be given. In Chapter 2 the methodology and results for Study 1 are outlined. In Chapter 3 the methodology and results for Study 2 are presented. Finally, in 4

16 Chapter 4 the results of this investigation are discussed as well as the studies limitations and impact on today s workplace. Background Attachment has been studied for four decades by various researchers using several psychological frameworks (i.e., clinical, development, cognitive, differential and social). Early theorists asserted that attachment theory attempted to explain the same phenomena as did theories of 'dependency needs', 'object relations', and 'individuation' (Ainsworth, 1969; Bowlby, 1980). In the remainder of this section I provide a brief review of both infant and romantic (i.e., close relationship) attachment theory. Overview John Bowlby's early ideas on personality development and its role in childprimary caregiver interactions stimulated substantial research on infant attachment (Bowlby, 1969/1982). Bowlby conceptualized attachment as a paradigm to explain the desire for human beings to make strong affectional bonds to others, such as primary caregivers; and to account for forms of emotional distress and personality disturbance (i.e., anger, emotional detachment, anxiety, and depression). Bowlby posited that unwilling separation and loss gave rise to emotional distress and personality disturbance factors in infants that are maintained throughout the individuals lives. Bowlby s attachment theory seeks to explain this phenomenon (Bowlby, 1980). Attachment theory proposes that human beings possess an innate disposition to direct attachment behaviors towards primary caregivers (Tracy, Lamb, & Ainsworth, 1976). Bowlby (1980) defined attachment behaviors as the following: 5

17 Attachment behavior is conceived as any form of behaviour that results in a person attaining or retaining proximity to some other differentiated and preferred individual. So long as the attachment figure remains accessible and responsive the behavior may consist of little more than checking by eye or ear on the whereabouts of the figure and exchanging occasional glances and greetings. In certain circumstances, however, following or clinging to the attachment figure may occur and also calling or crying, which are likely to elicit his or her caregiving. (p.39) These attachment behaviors form organized behavioral systems that have one thing in common, increasing proximity to a primary caregiver. Attachment behavior is a distinct concept from attachment bonds, which refer to "emotional ties that exist between individuals and their attachment figures" (Rholes & Simpson, 2004, p.4). In summary, Bowlby believed that primary caregivers could be described as a secure base from which an infant could explore the world and a sanctuary to come back to in times of stress. He claimed that continuous experience with the caregiver served as the foundation upon which individuals developed cognitive-affective structures about the self, others and interpersonal relationships. Bowlby termed these cognitive-affective structures "working models." "Working models," developed during the first year of life (Ainsworth, 1989), serve as shortcuts that enable infants to organize expectations about what happens to them. Developing and building upon Bowlby's original conceptualization of attachment theory, Mary Ainsworth pursued research on attachment theory in tandem with Bowlby. Ainsworth asserted that Bowlby's construct of attachment could be "elaborated and 6

18 redefined through further research" (Ainsworth et al., 1978, p. 4), and sought to develop an empirical test and classification for infant attachment. Ainsworth felt that "for Bowlby the inexplicable finding pertained to a young child's responses to separation from his mother figure" (Ainsworth et al, 1978, p. 4). Her research pursued the explanation of infant reactions to both separation and fear, which resulted in the classification of infants into three attachment styles: secure, anxious-ambivalent, and avoidant. Attachment styles are defined as: Stable, global individual differences in (1) tendencies to seek and experience comfort and emotional support from persons with whom one has an attachment bond and (2) presumptions about the responsiveness of attachment figures to bids for comfort and support. (Rholes & Simpson, 2004, p. 4) The attachment styles delineated by Ainsworth and her colleagues originated from the development of their test of infant attachment, the Strange Situation. The Strange-Situation procedure was based on Ainsworth and colleagues (1978) belief that an infant's response to separation from a primary caregiver evoked measurable attachment behaviors. The Strange Situation consisted of eight episodes in which infants experienced a series of distressing separations and reunions with their mothers and strangers. Early manipulation episodes were thought to be less stressful for the infant than later episodes. Infant behavior was monitored by an observer in an adjacent room through a one-way mirror. All sessions were either videotaped or a narrative record was made to determine differences in infants' behavior patterns. 7

19 Ainsworth and colleagues (1978) posited that differences in behavior patterns generated in the strange-situation episodes resulted from differences in infant and maternal behavior at home during the infant's first year of life (Ainsworth et al., 1978). Bowlby, as well as Ainsworth and colleagues, suggested distinct phases that the mother and infant go through that develop attachment. Therefore, each pattern of attachment behavior was thought to be associated with certain patterns of maternal care during the attachment phases (Ainsworth et al., 1978; Hazan & Shaver, 1987; Tracy & Ainsworth, 1981). Ainsworth and colleagues developed the Strange Situation to classify children at 12 and then 18 months into one of three attachment styles. This classificatory system (i.e., typology) was supplemented with quantitative measures (e.g., behavioral scores) to group infants depending on their patterns of behavior. Ainsworth and colleagues preferred the typology approach, as opposed to using only quantitative measures of behavior, because they believed it allowed for a more comprehensive view of behavior patterns and allowed for more flexibility with respect to new patterns that were not yet observed or quantified. The three main attachment categories developed using the strange-situation ratings were terms Grouped A, B, and C (Ainsworth et al., 1978); these groupings make-up the three main attachment styles that have been copiously studied over the past few decades. The most evident behavior exhibited by infants classified as belonging to the Group A category was avoidance of the mother in reunion episodes. This group of infants is commonly thought to have an avoidant attachment style. Avoidant infants frequently exhibit the behaviors Bowlby referred to as detachment. Subsequent research has suggested that approximately 23% of infants fall into this category (Campos, Barrett, 8

20 Lamb, Goldsmith, & Stenberg, 1983). Infants in the Group B classification sought proximity and contact with their mothers, especially during the reunion episodes. They tended to resist release from the mother and protest when the mother put them down. This group is posited to have a secure attachment style. Later research suggests that this group includes 62% of infants (Campos et al., 1983). The final category of infants, Group C, consisted of infants that demonstrated significant resistance from contact and interaction with the mother, especially during the last episode. Yet, these infants showed some seeking of proximity and contact to the mother and seeking to maintain contact once it was gained similar to a secure infant, which Ainsworth and colleagues believed made the infants seem ambivalent to the mother. Generally, this group displayed more maladaptive behaviors than other infants, such as excessive anger or passivity. This group of infants is often referred to as anxious-ambivalent; and infants in the anxiousambivalent category frequently exhibit the behaviors Bowlby called protest (Ainsworth et al., 1978). Research has shown that this is the smallest group of infants (this is consistent with Ainsworth and colleagues original findings), and this group is approximately 15% of the infant population (Campos et al., 1983). It is also noteworthy that Ainsworth and colleagues specified certain subgroups for all of the groups previously mentioned (A 1, A 2, B 1, B 2, B 3, B 4, C 1, C 2 ), but for this paper the specific subgroups will not be examined because they have not been frequently studied in the adult attachment literature. Adult Attachment Early conceptualizations of attachment asserted that infant attachment persists into adulthood (Ainsworth, 1989; Bowlby, 1969/1982). Attachment theory proposes that infant attachment experiences create "working models" and consequent behavior patterns 9

21 that are crucial to an individual's personality (Bowlby, 1973). There has been a lot of controversy regarding the continuity or stability of attachment, but there is research evidence to suggest continuity in the infant-primary caregiver attachment over time (Ainsworth, 1989; Lamb, Thompson, Gardner, Charnov, & Estes, 1984). For example, Main and colleagues (George, Kaplan, & Main, 1996; Main, Kaplan, & Cassidy, 1985) have done research on assessing attachment beyond infancy. They found that attachment can be measured in an individual's sixth year of life with similar results to that obtained during infancy using the strange-situation. Also, Hazan and Shaver (1987) found that adult attachment security and insecurity is predicted by one s self-reported attachment history. In addition, cross-generational patterns have been found (i.e., adults' attachment histories predict that of their children) implying there may also be continuity across generations. Generally, these continuity studies suggest that infant attachment has some stability across the lifespan and early attachment styles may influence childhood and adult behavior as well as attachment (George et al., 1996; Hazan & Shaver, 1987). Although attachment theorists believe that individuals form new principal attachment figures later in life (e.g., sexual pair bond), they maintain that the initial infant-parent attachment does not cease to exist (Ainsworth, 1989). More recent theories of adult attachment represent extensions of early childhood attachment theory. Attachment theory states that individuals may form new attachments later in life. Ainsworth (1989) made clear the distinctions among relationships, affectional bonds, and attachments. Relationships may include different behavioral systems, one of which is attachment. Affectional bonds are thought to be unique from relationships in that they are enduring, a characteristic of the internal organization of a 10

22 person (i.e., not dyadic like relationships), and they are more specific. Ainsworth claimed that attachments are unique affectional bonds in that they have all of the qualities of affectional bonds, but they also include comfort and security (i.e., a secure base) achieved through the relationship that allows the individual to explore or take on other activities. Theorists interested in adult attachment have interpreted Ainsworth's ideas on what attachment entails to expand the original theory to other contexts. Attachments are thought to be distinguished by: "(a) persistence, (b) specificity to a particular individual, (c) emotional significance, (d) desire for proximity or contact, (e) distress at involuntary separation, and (f) security and comfort seeking" (Cassidy, 2000, p. 111). Hazan & Shaver (1987, 1990) were the first to measure attachment using a target other than the primary caregiver inspiring research in social/personality psychology in which "attachment styles are conceptualized as systematic patterns of expectations, needs, emotions, and emotion-regulation strategies, and social behavior that result from the innate 'attachment behavioral system' (Bowlby, 1969/1982) and a particular history of attachment experiences" (Shaver & Mikulincer, 2002, p. 134). Hazan and Shaver viewed romantic relationships as attachment systems and examined the outcomes of individual differences in adult attachment. They suggested that romantic love is an attachment process that is distinct from infant-parent attachment, but is influenced by individuals' attachment histories given individual differences in the "working models" and behavior patterns created by initial attachment experiences. Hazan and Shaver's (1987) initial work on love as attachment mirrored Ainsworth and colleagues' work in that they assessed the same three attachment styles developed by the Strange Situation: avoidant, secure, and anxious-ambivalent. One 11

23 notable difference, however, was that Hazan and Shaver (1987) assessed the individual s attachment style though the use of a self-report questionnaire in which individuals selfclassified their most important romantic relationship into one of the three described styles of attachment. The authors conducted two studies, one in an academic context (i.e., undergraduate students) and one in which a love quiz was printed in a local newspaper. Individuals were asked to return the love quiz printed in the newspaper and the first 620 responses were used. Hazan and Shaver's methodology (i.e., a typology) will be discussed in greater detail later when I discuss the disadvantages of typologies. Hazan and Shaver (1987) found that 56% of the adults participating in the study classified themselves as secure, 24% classified themselves as avoidant, and finally 20% classified themselves as anxious-ambivalent. These findings were similar to the proportions found in infant-parent attachment studies (e.g., 62% secure, 23% avoidant, and 15% anxiousambivalent) demonstrating that the frequency of the attachment styles, even when directed toward a romantic partner rather than a primary caregiver, is somewhat stable from infancy to adulthood (Campos et al., 1983). These consistent findings served as initial evidence that the latent construct of attachment can be transferred to romantic relationships, which implies that various other domains may fulfill the criteria for an attachment bond. To further assess romantic attachment and its consistency with infant-parent attachment, Hazan and Shaver examined the adult participants' romantic experiences, "working models," attachment histories (from memory), and feelings of loneliness. Their findings indicated that the adult attachment styles were associated with different emotions regarding romantic experiences in that each of the attachment styles were tied 12

24 to a different set of emotions (i.e., the emotion means were significantly different for the three attachment styles). Secure individuals experienced more trust (academic sample means for trust: secure M = 3.57, anxious M = 3.35, and avoidant M = 3.25, F(2, 104) = 3.03 p <.05; newspaper sample means for trust: secure M = 3.43, anxious M = 3.13, and avoidant M = 3.11, F(2, 571) = 16.21, p <.001), friendship (academic sample means for friendship: secure M = 3.61, anxious M = 3.39, and avoidant M = 3.34, F(2, 104) = 3.30, p <.05; newspaper sample means for friendship: secure M = 3.50, anxious M = 3.19, and avoidant M = 3.18, F(2, 571) = 22.96, p <.001), and less fear of closeness in love experiences (academic sample means for fear of closeness: secure M = 2.13, anxious M = 2.45, and avoidant M = 2.63, F(2, 104) = 4.48, p <.01; newspaper sample means for fear of closeness: secure M =1.88, anxious M = 2.15, and avoidant M = 2.30, F(2, 571) = 22.65, p <.001). Avoidant individuals had higher levels of fear of closeness than secure and anxious-ambivalent individuals. Finally, anxious-ambivalent individuals reported higher levels of desire for union (academic sample means for desire for union: anxious M = 3.29, avoidant M = 2.83, and secure M = 2.92, F(2, 104) = 3.41, p <.05; newspaper sample means for desire for union: anxious M = 3.25, avoidant M = 2.81, and secure M = 2.69, F(2, 571) = 22.67, p <.001), and desire for reciprocation (academic sample means for desire for reciprocation: anxious M = 3.64, avoidant M = 3.21, and secure M = 3.18, F(2, 104) = 7.50, p <.001; newspaper sample means for desire for reciprocation: anxious M = 3.55, avoidant M = 3.24, and secure M = 3.22, F(2, 571) = 14.90, p <.001). These findings regarding attachment differences in adults feelings experienced in romantic relationships are consistent with conceptualizations of infant attachment in that 13

25 attachment security is related to the infant's feelings experienced in his or her relationship with a primary caregiver. Additionally, Hazan and Shaver found that the adult's "working models" of self and relationships were related to their attachment style, in that mean differences were found among the attachment styles regarding agreement with "mental-model" statements. For example, anxious-ambivalent individuals were significantly more likely to agree with the statement "It's easy to fall in love. I feel myself beginning to fall in love often" (Hazan & Shaver, 1987, p. 516) than secure and avoidant individuals (i.e., avoidant M =.04, anxious M =.20, and secure M =.09, F(2, 571) = 9.33, p <.001). These differences suggest that the three attachment styles are associated with different expectations about romantic love and beliefs about others, and therefore that the "working models" in Bowlby's infant attachment theory are also applicable in adult attachment. Further evidence to support attachment consistency from infancy to adulthood was partially demonstrated in Hazan and Shaver's study through the prediction of secure versus insecure attachment styles via attachment histories. This was done using adjective checklists regarding the individual's relationship with his or her parent and his or her parent's marital relationship. Hazan and Shaver's (1987) extension of attachment into the adult domain served as a catalyst for a whole new area of research. There have been many subsequent studies on adult attachment demonstrating small to medium correlations between romantic/close relationship attachment and personality as well as interpersonal traits. All correlation effect sizes in this paper will follow the standard for size of effects for correlation coefficients in Cohen (1988); effects from r =.10 to r =.29 are considered small, r =.30 to r =.49 are considered medium and those over r =.50 are 14

26 considered large (e.g., Shaver & Brennan, 1992; Gallo et al., 2003). Also, moderate associations have been shown between self-report romantic attachment patterns and behavior within romantic relationships both reported by the romantic partners and observed within the laboratory setting (Jacobvitz et al., 2002; Simpson, Rholes, & Nelligan, 1992; Simpson, Rholes, & Phillips, 1996). Almost all of these studies have maintained that romantic attachment is the adult parallel to Bowlby and Ainsworth's infant-parent attachment (Shaver & Mikulincer, 2004). Attachment and Work Attachment theory (Bowlby, 1969/1982) has not commonly been studied by industrial and organizational psychologists. However, attachment theory provides a potentially useful framework for understanding the ties between personality, affect and motivation, which are individual differences constructs that are critical to organizational theories (Barrick & Mount, 1991; Gallo et al., 2003; Shaver & Mikulincer, 2002). Attachment theory presents potentially useful constructs for understanding the overlap in these individual differences variables. The nomological network of attachment and its relationship with motivational and personality traits will be discussed in more detail in later sections. In contrast to the social and personality domains, theories of attachment found in the organizational literature have taken a very macro (i.e., systems-oriented) perspective and have largely ignored dispositional factors (i.e., person-focused factors) of employee attachment (Herscovitch & Meyer, 2002). Additionally, research on the prediction of interpersonal behaviors or "relationship skills," which are associated with the attachment patterns, has largely ignored attachment theory and opted for more 15

27 common, broad personality dimensions frequently used in the organizational literature such as the Big Five (e.g., Mount et al., 1998). As previously discussed, attachment in the organizational research has largely been viewed as an employee attitude towards a situation or impersonal target (e.g., Herscovitch & Meyer, 2002; Mathieu & Zajac, 1990; Meyer & Allen, 1991). Examples of these attitudinal theories related to workers' psychological states are: organizational commitment (i.e., the likelihood that an individual will stay in a certain organization), job satisfaction (i.e., an affective and cognitive evaluation of one's job), and job embeddedness (Brief & Weiss, 2002; Meyer & Allen, 1991; Mitchell et al., 2001). These variables are largely thought to be improved or enhanced by system-oriented processes such as organizational socialization (Bauer, Morrison, & Callister, 1998), and they serve as crucial constructs for understanding turnover within the organization (Mitchell et al., 2001). Yet, individual differences are predictive of these psychological states in addition to the macro-level interventions. Research on job satisfaction has demonstrated that biologically based traits influence one s level of job satisfaction and 30% of the variance in job satisfaction can be attributed to genetics (Arvey, Bouchard, Segal, & Abraham, 1989). Some individuals inherit dispositions that lead them to seek happiness while other individuals are disposed to seek negative situations (Arvey et al., 1989; Staw & Ross, 1989). Despite limited areas of research on dispositions such as genetics and job satisfaction, individual differences related to these "attachment" constructs have been less prominent in the literature than the environment in which the individual works. Organizational research on attachment to people within the organization and interpersonal relations is crucial for today's service-based industry. Research in this area 16

28 includes studies on interpersonal behaviors (i.e., relationship skills), contextual performance (e.g., cooperative behaviors), and various other variables, which typically focus on macro-level predictors or broad personality traits, as opposed to individual characteristics that influence interpersonal functioning (Borman & Motowidlo, 1997; Kanfer, 2003; Lepine & Van Dyne, 2001; Mount et al., 1998). Attachment in the social domain is conceptualized as an individual-level disposition that guides behaviors and emotions with respect to others, and therefore it is reasonable to assume that attachment would predict these important criteria (Feeney, 1999; Feeney, 2004; Shaver & Mikulincer, 2002). Therefore, worker/peer attachment presented a potentially fruitful variable that taps the dyadic nature of relationships within an organization and predicts employee interpersonal or relationship-based performance and work attitudes. Despite this disconnect between attachment in the work domain and attachment in development/social psychology, there is limited empirical work investigating the relationship between romantic attachment and work-related variables. The original work was conducted by Hazan & Shaver (1990) using the adult romantic attachment styles they developed (Hazan & Shaver, 1987) and measures of work-related criteria such as job satisfaction and demographic variables. The authors suggested that work is similar to what Bowlby (1969/1982) called exploration in that romantic adult attachment supports work behaviors in the same fashion that infant attachment supports exploration. Given this parallel the authors hypothesized the following: In the same way that Ainsworth et al.'s (1978) avoidant infants appeared to explore to avoid seeking contact with their mothers, adults can approach their work compulsively or use it as a distraction from relational 17

29 deficiencies. For someone with anxious/ambivalent proclivities, work can be viewed as an opportunity to satisfy attachment needs, a sideline that may interfere with job performance (Hazan & Shaver, 1990, p. 271). The results from Hazan and Shaver's study were in the expected direction given the beneficial affect-regulation strategies and motivational tendencies of secure individuals (Elliot and Reis, 2003; Shaver & Mikulincer, 2002). Securely attached individuals had higher levels of work satisfaction in terms of job security, coworkers, income, opportunities for challenge, and advancement. They were more confident about their work and were not consumed by fear of failure (i.e., they had lower means for worry they won't impress others and fear of rejection for poor work than insecure individuals: secure M = 2.29, avoidant M = 2.83, and anxious M = 2.88, F(2, 233) = 10.98, p <.001). Additionally, secure individuals reported that they valued work, but they viewed their relationships as more important than work (means for relationship more important than work: secure M = 1.81, avoidant M = 1.62, anxious M = 1.67, F(2, 658) = 12.34, p <.001), and they did not let problems from work interfere with their relationships (means for work harms health/relationships: avoidant M = 2.50, anxiety M = 2.25, and secure M = 1.93, F(2, 233) = 12.31, p <.001). Anxious-ambivalent individuals had lower levels of work satisfaction concerning job security (means for happy with job security: anxiety M = 2.76, avoidant M =2.76, and secure M = 2.91, F(2, 658) = 6.69, p <.001), coworker recognition (means for unhappy with recognition: anxiety M = 2.36, avoidant M = 2.18, and secure M = 1.99, F(2, 658) = 6.42, p <.001), and advancement (means for unhappy with advancement: anxiety M = 2.34, avoidant M = 2.22, and secure M = 2.08, F(2, 658) = 3.08, p <.05). 18

30 These individuals reported that problems from their relationships often interfere with their work performance (i.e., poor affect-regulation) and they are afraid of failure due to the rejection it may bring (Hazan & Shaver, 1990). Assuming attachment is a disposition developed in infancy, anxious individuals' motivation is impaired by their fear of failure and therefore they create avoidance-related goals. This is partially demonstrated with Hazan and Shaver s (1990) finding that anxious individuals had the lowest incomes of the three attachment styles. The average means for level of income were $30,000 to $40,000 in the sample; whereas the anxiety income mean was $20,000 to $30,000, and the secure and avoidant income means were the same as the average (i.e., the secure and avoidance income means were significantly higher than the anxiety mean F(2, 644) = 24.83, p <.001). The avoidant attachment style was related to low levels of satisfaction with coworkers, but it was similar to secure attachment in its relationship to satisfaction with job security and opportunities for learning. Also, avoidant individuals reported comparable incomes to secure individuals, but they were less likely to want to take vacations (means for viewing vacations are pleasureless: avoidant M = 2.55, anxiety M = 2.37, and secure M = 2.13, F(2, 233) = 5.79, p <.01), and they felt that work interfered with social activities (means for work harms health/relationships: avoidant M = 2.50, anxiety M = 2.25, and secure M = 1.93, F(2, 233) = 12.31, p <.001). That is, avoidant individuals were proposed to use work as an excuse to avoid social situations (Hazan & Shaver, 1990). Despite Hazan and Shaver s (1990) methodology (i.e., using categorical, single-item instruments) these initial results provide evidence that there is a relationship between attachment and work-related variables given the significantly different means 19

31 found for work-related variables among the three attachment styles. The current investigation provides a lot more information about these relationships. Two other studies investigating the linkage between romantic attachment and work have examined individual differences in attachment (i.e., attachment anxiety and avoidance traits) as inputs to team-related processes and outcomes (Rom and Mikulincer, 2003; Smith, Murphy, and Coats, 1999). Smith and colleagues (1999) developed a measure of group attachment by modifying existing self-report anxiety and avoidance attachment scales to refer to social groups as opposed to a romantic partner. They found that group attachment predicts group identification (anxiety r = -.32, p <.001, avoidance r = -.75, p <.001) and a variety of group processes and outcomes, such as: negative affect (anxiety r =.46, p <.001, avoidance r =.28, p <.001), the amount of activity an individual gives to a group (anxiety r = -.31, p <.05, avoidance r =.15, ns) time they spend with the group (anxiety r = -.27, p <.001, avoidance r = -.11, ns), number of social supports (anxiety r = -.31, p <.001, avoidance r = -.56, p <.001) satisfaction with social supports (anxiety r = -.44, p <.001, avoidance r = -.59, p <.001), collective selfesteem (anxiety r = -.48, p <.001, avoidance r = -.45, p <.001), and ways of resolving conflicts (Smith et al., 1999). Attachment avoidance was negatively related to positive affect (r = -.58, p <.001) and positively correlated with fewer and less satisfying social supports (r = -.56, p <.001) and plans to leave the group (r =.65, p <.001). It is interesting to note that group attachment was found to be conceptually and empirically distinct from adult romantic attachment (correlations between group and relationship attachment: group attachment anxiety and relationship attachment anxiety r =.70, group 20

32 attachment avoidance and relationship attachment avoidance r =.41). This suggests a degree of independence between group and romantic relationship attachment orientations. If individuals create different attachment patterns in different contexts with different targets, it is possible that individuals with secure romantic attachment may have insecure attachment related to work. Yet, previous attachment experiences will cause some amount of consistency from prior attachment styles to attachment in the work domain. This is similar to the consistency from infant to romantic/close relationship attachment (Cassidy, 2000), or romantic/close relationship attachment to group attachment (Smith et al., 1999). There should be consistency in romantic/close relationship attachment and worker attachment in that close relationship attachment will be an important antecedent to worker attachment. Basically, one's infant attachment style, or "working models" of self and others, carries across the lifespan (Bowlby 1969/1982) influencing romantic/close relationship attachment, group attachment, (Cassidy, 2000; Shaver & Mikulincer, 2004; Smith et al., 1999) and worker/peer attachment. Because individuals are posited to develop relationships with others prior to their initial work experience, it seems reasonable to propose that romantic/close relationship attachment affects worker attachment. Yet, similar to romantic/close relationship attachment, extreme or pivotal situations can change one's attachment style (Cassidy, 2000; Shaver & Mikulincer, 2002). In the workplace, socialization and early work experiences, where an individual is developing his/her model of self and others as a worker, may provide such a context for changes in attachment. A follow-up study to the work done by Smith and colleagues (1999) was conducted by Rom and Mikulincer (2003). The authors claimed that attachment theory is 21

33 relevant to group contexts given the group as a whole or individual group members can fulfill the "definitional criteria" for what constitutes an attachment bond (i.e. proximity seeking, support and relief during times of stress, and exploration facilitation), suggesting that individual differences in attachment will influence affective reactions, cognition, and behavior in group interactions. They proposed that the study done by Smith et al. (1999) did not measure some of the important variables for understanding the link between group processes and the attachment system such as cognition, affect, and behavior during group interactions. Additionally, Smith et al.'s work did not investigate any group-level variables such as group cohesion, which is defined as "the level of coordination, cooperation, support, and consensus that exist among group members (Hogg, 1992; Levine & Moreland, 1990)" (Rom & Mikulincer, 2003). Rom and Mikulincer posited that group cohesion could be viewed as a part of the group attachment bond since group members in highly cohesive groups would receive more support from the group during times of stress. Finally, the authors measured attachment working models that were not considered by Smith and colleagues. As expected, adult attachment anxiety and avoidance were found to predict negative group-related cognitions, such as challenge-appraisal (avoidance β = -.36, p <.01) and threat appraisal (anxiety β =.43, p <.01), and emotions (anxiety β =.34, p <.01, avoidance β =.28, p <.01) due to the hyperactivating and deactivating strategies used by insecure individuals that will be discussed in more detail later. Attachment anxiety was indicative of a negative model of the self as a group member, which was claimed to be a consequence of their overall negative model of self in which they see themselves as helpless and unworthy. Anxious individuals saw group interactions as a threat and had 22